Replay and rollback attacks threaten cloud application integrity by reintroducing authentic yet stale data through an untrusted storage interface to compromise application decision-making. Prior security frameworks mitigate these attacks by enforcing forward-only state transitions (state continuity) with hardware-backed mechanisms, but they categorically treat all rollback as malicious and thus preclude legitimate rollbacks used for operational recovery from corruption or misconfiguration. We present Rebound, a general-purpose security framework that preserves rollback protection while enabling policy-authorized legitimate rollbacks of application binaries, configuration, and data. Key to Rebound is a reference monitor that mediates state transitions, enforces authorization policy, guarantees atomicity of state updates and rollbacks, and emits a tamper-evident log that provides transparency to applications and auditors. We analyze Rebound's security properties and show through an application case study -- with software deployment workflows in GitLab CI -- that it enables robust control over binary, configuration, and raw data versioning with low end-to-end overhead.

Texture data serves as a valuable tool for interpreting the high-level features models learn, uncovering biases, and identifying security vulnerabilities. However, works in this space have been limited by small texture datasets and synthesis methods that struggle to scale in the diversity and specificity required for these tasks. In this work, we introduce an extensible methodology for generating high-quality, diverse texture images, which we use to create the Prompted Textures Dataset (PTD), a new texture dataset spanning 246,285 images across 56 texture classes. Our comparison against real texture data demonstrates that PTD is more diverse while maintaining quality. Additionally, human evaluations confirm that every stage in our methodology enhances texture quality, yielding a 3.4% increase in quality and a 4.5% increase in representativeness overall. Our dataset is available for download at https://zenodo.org/records/15359142.

Fine-tuning has become the standard practice for adapting pre-trained models to downstream tasks. However, the impact on model robustness is not well understood. In this work, we characterize the robustness-accuracy trade-off in fine-tuning. We evaluate the robustness and accuracy of fine-tuned models over 6 benchmark datasets and 7 different fine-tuning strategies. We observe a consistent trade-off between adversarial robustness and accuracy. Peripheral updates such as BitFit are more effective for simple tasks -- over 75% above the average measured by the area under the Pareto frontiers on CIFAR-10 and CIFAR-100. In contrast, fine-tuning information-heavy layers, such as attention layers via Compacter, achieves a better Pareto frontier on more complex tasks -- 57.5% and 34.6% above the average on Caltech-256 and CUB-200, respectively. Lastly, we observe that the robustness of fine-tuning against out-of-distribution data closely tracks accuracy. These insights emphasize the need for robustness-aware fine-tuning to ensure reliable real-world deployments.

Storage integrity is essential to systems and applications that use untrusted storage (e.g., public clouds, end-user devices). However, known methods for achieving storage integrity either suffer from high (and often prohibitive) overheads or provide weak integrity guarantees. In this work, we demonstrate a hybrid approach to storage integrity that simultaneously reduces overhead while providing strong integrity guarantees. Our system, partially asynchronous integrity checking (PAC), allows disk write commitments to be deferred while still providing guarantees around read integrity. PAC delivers a 5.5 × throughput and latency improvement over the state of the art, and 85% of the throughput achieved by non-integrity-assuring approaches. In this way, we show that untrusted storage can be used for integrity-critical workloads without meaningfully sacrificing performance.

Cloud file systems offer organizations a scalable and reliable file storage solution. However, cloud file systems have become prime targets for adversaries, and traditional designs are not equipped to protect organizations against the myriad of attacks that may be initiated by a malicious cloud provider, co-tenant, or end-client. Recently proposed designs leveraging cryptographic techniques and trusted execution environments (TEEs) still force organizations to make undesirable trade-offs, consequently leading to either security, functional, or performance limitations. In this paper, we introduceBFS, a cloud file system that leverages the security capabilities provided by TEEs to bootstrap new security protocols that deliver strong security guarantees, high-performance, and a transparent POSIX-like interface to clients.BFSdelivers stronger security guarantees and up to a$2.5\times$speedup over a state-of-the-art secure file system. Moreover, compared to the industry standard NFS,BFSachieves up to$2.2\times$speedups across micro-benchmarks and incurs$< 1\times$overhead for most macro-benchmark workloads.BFSdemonstrates a holistic cloud file system design that does not sacrifice an organizations’ security yet can embrace all of the functional and performance advantages of outsourcing.

Modern host-based intrusion detection systems (HIDS) rely on querying provenance graphs—graph representations of activity history on a system—to detect and respond to security threats present on a system. However, as the complexity and number of applications running on a system increase, the size of provenance graphs also increase, and thus the latency to query them. State-of-the-art designs deliver query latencies that are impractical for modern threat detection. In this paper, we introduce a hyper-dimensional computing (HDC) approach to querying provenance graphs for HIDS. By encoding provenance graphs and attack patterns/signatures into hyper-dimensional vectors, we can implement a query engine using simple vector operations. Our approach is hardware accelerator compatible, providing further speedups under resource-constrained environments. Our evaluation on a real-world dataset shows that our approach achieves > 90% detection accuracy and up to 4, 242× speedups over the state-of-the-art. This shows that HDC-based approaches can effectively deal with scaling issues in modern HIDS.

Today, targeted online advertising relies on unique identifiers assigned to users through third-party cookies--a practice at odds with user privacy. While the web and advertising communities have proposed solutions that we refer to as interest-disclosing mechanisms, including Google's Topics API, an independent analysis of these proposals in realistic scenarios has yet to be performed. In this paper, we attempt to validate the privacy (i.e., preventing unique identification) and utility (i.e., enabling ad targeting) claims of Google's Topics proposal in the context of realistic user behavior. Through new statistical models of the distribution of user behaviors and resulting targeting topics, we analyze the capabilities of malicious advertisers observing users over time and colluding with other third parties. Our analysis shows that even in the best case, individual users' identification across sites is possible, as 0.4% of the 250k users we simulate are re-identified. These guarantees weaken further over time and when advertisers collude: 57% of users with stable interests are uniquely re-identified when their browsing activity has been observed for 15 epochs, increasing to 75% after 30 epochs. While measuring that the Topics API provides moderate utility, we also find that advertisers and publishers can abuse the Topics API to potentially assign unique identifiers to users, defeating the desired privacy guarantees. As a result, the inherent diversity of users' interests on the web is directly at odds with the privacy objectives of interest-disclosing mechanisms; we discuss how any replacement of third-party cookies may have to seek other avenues to achieve privacy for the web.

Load balancing in software-defined networks (SDNs) is commonly realized with a centralized architecture. Dynamic load balancing relies on the SDN controller to periodically collect traffic statistics from network switches and make decisions in a timely manner. In this paper, we examine the extent to which an adversary that has compromised a switch can influence the load balancing algorithm by misreporting its own traffic statistics. We design an attack that allows an adversary to perform preliminary reconnaissance, which means learning network traffic distributions and setting attack parameters, and then accurately model and estimate the reward from misreporting while evading detection. Our evaluation offers three insights: 1) network traffic exhibits discernible patterns by reconnaissance; 2) the reconnaissance can be used to design misreporting attacks that can effectively draw unfair proportions of network traffic to the adversary under the guise of honest behavior; and 3) reconnaissance itself can be accelerated by misreporting to launch more targeted attacks.

As a novel network paradigm, Software Defined Networking (SDN) has greatly simplified network management, but also introduced new vulnerabilities. One vulnerability of particular interest is the flow table, a data structure in every SDN-enabled switch that caches flow rules from the controller to bridge the speed gap between the data plane and the control plane. Prior works have shown that an adversary-controlled host can accurately infer parameters of the flow table at its directly-connected edge switch, which can then be used to launch intelligent attacks. However, those solutions do not work for flow tables at internal switches. In this work, we develop an algorithm that can infer the different flow table sizes at internal switches by measuring the Round Trip Times (RTTs) of a path traversing these switches from one of its endpoints. A major challenge in this problem is the lack of an inferable relationship between the RTTs and the flow table hits/misses at the traversed switches. Our solution addresses this challenge by experimentally identifying the inferable information and designing an inference algorithm that combines carefully designed probing sequences and statistical tools to mitigate measurement noise and interference. The efficacy of our solution is validated through experiments in Mininet.

Software security depends on coordinated vulnerability disclosure (CVD) from researchers, a process that the community has continually sought to measure and improve. Yet, CVD practices are only as effective as the data that informs them. In this paper, we use DScope, a cloud-based interactive Internet telescope, to build statistical models of vulnerability lifecycles, bridging the data gap in over 20 years of CVD research. By analyzing application-layer Internet scanning traffic over two years, we identify real-world exploitation timelines for 63 threats. We bring this data together with six additional datasets to build a complete birth-to-death model of these vulnerabilities, the most complete analysis of vulnerability lifecycles to date. Our analysis reaches three key recommendations: (1) CVD across diverse vendors shows lower effectiveness than previously thought, (2) intrusion detection systems are underutilized to provide protection for critical vulnerabilities, and (3) existing data sources of CVD can be augmented by novel approaches to Internet measurement. In this way, our vantage point offers new opportunities to improve the CVD process, achieving a safer software ecosystem in practice.

In-network caching and flexible routing are two of the most celebrated advantages of next generation network infrastructures. Yet few solutions are available for jointly optimizing caching and routing that provide performance guarantees for networks with arbitrary topology. We take a holistic approach towards this fundamental problem by analyzing its complexity in all the cases and developing polynomial-time algorithms with approximation guarantees in important special cases. We also reveal the fundamental challenge in achieving guaranteed approximation in the general case and propose an alternating optimization algorithm with good empirical performance and fast convergence. Our algorithms have demonstrated superior performance in both routing cost and congestion compared to the state-of-the-art solutions in evaluations based on real topology and request traces.

Similarities between natural languages and programming languages have prompted researchers to apply neural network models to software problems, such as code generation and repair. However, program-specific characteristics pose unique prediction challenges that require the design of new and specialized neural network solutions. In this work, we identify new prediction challenges in application programming interface (API) completion tasks and find that existing solutions are unable to capture complex program dependencies in program semantics and structures. We design a new neural network model Multi-HyLSTM to overcome the newly identified challenges and comprehend complex dependencies between API calls. Our neural network is empowered with a specialized dataflow analysis to extract multiple global API dependence paths for neural network predictions. We evaluate Multi-HyLSTM on 64,478 Android Apps and predict 774,460 Java cryptographic API calls that are usually challenging for developers to use correctly. Our Multi-HyLSTM achieves an excellent top-1 API completion accuracy at 98.99%. Moreover, we show the effectiveness of our design choices through an ablation study and have released our dataset.

No abstract available.

In-network caching and flexible routing are two of the most celebrated advantages of next generation network infrastructures. Yet few solutions are available for jointly optimizing caching and routing that provide performance guarantees for an arbitrary topology. We take a holistic approach towards this fundamental problem by analyzing its complexity in all the cases and developing polynomial-time algorithms with approximation guarantees in important special cases. We also reveal the fundamental challenge in achieving guaranteed approximation in the general case and propose an alternating optimization algorithm with good performance and fast convergence. Our algorithms have demonstrated superior performance in both routing cost and congestion compared to the state-of-the-art solutions in evaluations based on real topology and request traces.

Public clouds provide scalable and cost-efficient computing through resource sharing. However, moving from traditional on-premises service management to clouds introduces new challenges; failure to correctly provision, maintain, or decommission elastic services can lead to functional failure and vulnerability to attack. In this paper, we explore a broad class of attacks on clouds which we refer to as cloud squatting. In a cloud squatting attack, an adversary allocates resources in the cloud (e.g., IP addresses) and thereafter leverages latent configuration to exploit prior tenants. To measure and categorize cloud squatting we deployed a custom Internet telescope within the Amazon Web Services us-east-1 region. Using this apparatus, we deployed over 3 million servers receiving 1.5 million unique IP addresses ($\approx$ 56% of the available pool) over 101 days beginning in March of 2021. We identified 4 classes of cloud services, 7 classes of third-party services, and DNS as sources of exploitable latent configurations. We discovered that exploitable configurations were both common and in many cases extremely dangerous; we received over 5 million cloud messages, many containing sensitive data such as financial transactions, GPS location, and PII. Within the 7 classes of third-party services, we identified dozens of exploitable software systems spanning hundreds of servers (e.g., databases, caches, mobile applications, and web services). Lastly, we identified 5446 exploitable domains panning 231 eTLDs—including 105 in the top 10000 and 23 in the top 1000 popular domains. Through tenant disclosures we have identified several root causes, including (a) a lack of organizational controls, (b) poor service hygiene, and (c) failure to follow best practices. We conclude with a discussion of the space of possible mitigations and describe the mitigations to be deployed by Amazon in response to this study.

Machine learning-based network intrusion detection systems have demonstrated state-of-the-art accuracy in flagging malicious traffic. However, machine learning has been shown to be vulnerable to adversarial examples, particularly in domains such as image recognition. In many threat models, the adversary exploits the unconstrained nature of images–the adversary is free to select some arbitrary amount of pixels to perturb. However, it is not clear how these attacks translate to domains such as network intrusion detection as they contain domain constraints, which limit which and how features can be modified by the adversary. In this paper, we explore whether the constrained nature of networks offers additional robustness against adversarial examples versus the unconstrained nature of images. We do this by creating two algorithms: (1) the Adapative-JSMA, an augmented version of the popular JSMA which obeys domain constraints, and (2) the Histogram Sketch Generation which generates adversarial sketches: targeted universal perturbation vectors that encode feature saliency within the envelope of domain constraints. To assess how these algorithms perform, we evaluate them in a constrained network intrusion detection setting and an unconstrained image recognition setting. The results show that our approaches generate misclassification rates in network intrusion detection applications that were comparable to those of image recognition applications (greater than 95%). Our investigation shows that the constrained attack surface exposed by network intrusion detection systems is still sufficiently large to craft successful adversarial examples – and thus, network constraints do not appear to add robustness against adversarial examples. Indeed, even if a defender constrains an adversary to as little as five random features, generating adversarial examples is still possible.

Abstract Millions of consumers depend on smart camera systems to remotely monitor their homes and businesses. However, the architecture and design of popular commercial systems require users to relinquish control of their data to untrusted third parties, such as service providers (e.g., the cloud). Third parties therefore can (and in some instances have) access the video footage without the users’ knowledge or consent—violating the core tenet of user privacy. In this paper, we present CaCTUs, a privacy-preserving smart Camera system Controlled Totally by Users. CaCTUs returns control to the user; the root of trust begins with the user and is maintained through a series of cryptographic protocols, designed to support popular features, such as sharing, deleting, and viewing videos live. We show that the system can support live streaming with a latency of 2 s at a frame rate of 10 fps and a resolution of 480 p. In so doing, we demonstrate that it is feasible to implement a performant smart-camera system that leverages the convenience of a cloud-based model while retaining the ability to control access to (private) data.

IoT devices can be used to complete a wide array of physical tasks, but due to factors such as low computational resources and distributed physical deployment, they are susceptible to a wide array of faulty behaviors. Many devices deployed in homes, vehicles, industrial sites, and hospitals carry a great risk of damage to property, harm to a person, or breach of security if they behave faultily. We propose a general fault handling system named IoTRepair, which shows promising results for effectiveness with limited latency and power overhead in an IoT environment. IoTRepair dynamically organizes and customizes fault-handling techniques to address the unique problems associated with heterogeneous IoT deployments. We evaluate IoTRepair by creating a physical implementation mirroring a typical home environment to motivate the effectiveness of this system. Our evaluation showed that each of our fault-handling functions could be completed within 100 milliseconds after fault identification, which is a fraction of the time that state-of-the-art fault-identification methods take (measured in minutes). The power overhead is equally small, with the computation and device action consuming less than 30 milliwatts. This evaluation shows that IoTRepair not only can be deployed in a physical system, but offers significant benefits at a low overhead.

Multiple users have access to multiple devices in a smart home system – typically through a dedicated app installed on a mobile device. Traditional access control mechanisms consider one unique, trusted user that controls access to the devices. However, multi-user multi-device smart home settings pose fundamentally different challenges to traditional single-user systems. For instance, in a multi-user environment, users have conflicting, complex, and dynamically-changing demands on multiple devices that cannot be handled by traditional access control techniques. Moreover, smart devices from different platforms/vendors can share the same home environment, making existing access control obsolete for smart home systems. To address these challenges, in this paper, we introduce Kratos+ , a novel multi-user and multi-device-aware access control mechanism that allows smart home users to flexibly specify their access control demands. Kratos+ has four main components: user interaction module, backend server, policy manager, and policy execution module. Users can easily specify their desired access control settings using the interaction module that are translated into access control policies in the back-end server. The policy manager analyzes these policies, initiates automated negotiation between users to resolve conflicting demands, and generates final policies to enforce in smart home systems. We implemented Kratos+ as a platform-independent solution and evaluated its performance on real smart home deployments featuring multi-user scenarios with a rich set of configurations (337 different policies including 231 demand conflicts and 69 restriction policies). These configurations also included five different threats associated with access control mechanisms. Our extensive evaluations show that Kratos+ is very effective in resolving conflicting access control demands with minimal overhead. We also performed an extensive user study with 72 smart home users to better understand the user’s needs before designing the system and a usability study to evaluate the efficacy of Kratos+ in a real-life smart home environment.

Multilevel security (MLS) systems control access to data by formalizing permissible and impermissible information flows between data sources and destinations (e.g., database servers and clients) fixed with distinct security labels. In computer networks, MLS systems have been used to prevent unauthorized data disclosure in shared-infrastructure settings where network hosts and devices may fall within different trust domains (e.g., in multi-tenant cloud networks or wireless mesh networks). However, current MLS systems assume static network behavior—thus preventing the network from being practically usable in the presence of dynamic network events that frequent unstable network environments, including sudden changes in traffic patterns, link failures, and topology changes as a result of device movement or intermittent device connectivity. In this paper, we introduceMLS-Enforcer, a software-defined networking (SDN) controller application that can efficiently deploy network-level MLS policies while retaining the ability to securely relabel network nodes under changing topology state and network traffic demands. We model network adaptivity as an integer linear programming problem that reflects a given security policy. We then introduce heuristic relabeling algorithms that achieve near-optimal performance and are more tractable and efficient for larger networks. We validateMLS-Enforceron several network topologies and traffic loads, demonstrating that it can relabel the network to route 90%+ of flows under normal conditions and quickly converge (on the order of seconds for the heuristic algorithms) under changing needs—from small network structure changes to catastrophic failures. This shows that formally secured networks can feasibly be deployed in diverse, changing, and unpredictable environments.

Caches are pervasively used in communication networks to speed up content access by reusing previous communications, where various replacement policies are used to manage the cached contents. The replacement policy of a cache plays a key role in its performance, and is thus extensively engineered to achieve a high hit ratio in benign environments. However, some studies showed that a policy with a higher hit ratio in benign environments may be more vulnerable to cache pollution attacks that intentionally send requests for unpopular contents. To understand the cache performance under such attacks, we analyze a suite of representative replacement policies under the framework of TTL approximation in how well they preserve the hit ratios for legitimate users, while incorporating the delay for the cache to obtain a missing content. We further develop a scheme to adapt the cache replacement policy based on the perceived level of attack. Our analysis and validation on real traces show that although no single policy is resilient to all the attack strategies, suitably adapting the replacement policy can notably improve the attack resilience of the cache. Motivated by these results, we implement selected policies as well as policy adaptation in an open-source SDN switch to manage flow rule replacement, which is shown to notably improve its resilience to pollution attacks.

Machine learning is vulnerable to adversarial examples--inputs designed to cause models to perform poorly. However, it is unclear if adversarial examples represent realistic inputs in the modeled domains. Diverse domains such as networks and phishing have domain constraints--complex relationships between features that an adversary must satisfy for an attack to be realized (in addition to any adversary-specific goals). In this paper, we explore how domain constraints limit adversarial capabilities and how adversaries can adapt their strategies to create realistic (constraint-compliant) examples. In this, we develop techniques to learn domain constraints from data, and show how the learned constraints can be integrated into the adversarial crafting process. We evaluate the efficacy of our approach in network intrusion and phishing datasets and find: (1) up to 82% of adversarial examples produced by state-of-the-art crafting algorithms violate domain constraints, (2) domain constraints are robust to adversarial examples; enforcing constraints yields an increase in model accuracy by up to 34%. We observe not only that adversaries must alter inputs to satisfy domain constraints, but that these constraints make the generation of valid adversarial examples far more challenging.

Caches are pervasively used in computer networks to speed up access by reusing previous communications, where various replacement policies are used to manage the cached contents. The replacement policy of a cache plays a key role in its performance, and is thus extensively engineered to achieve a high hit ratio in benign environments. However, some studies showed that a policy with a higher hit ratio in benign environments may be more vulnerable to denial of service (DoS) attacks that intentionally send requests for unpopular contents. To understand the cache performance under such attacks, we analyze a suite of representative replacement policies under the framework of TTL approximation in how well they preserve the hit ratios for legitimate users, while incorporating the delay for the cache to obtain a missing content. We further develop a scheme to adapt the cache replacement policy based on the perceived level of attack. Our analysis and validation on real traces show that although no single policy is resilient to all the attack strategies, suitably adapting the replacement policy can notably improve the attack resilience of the cache.

Abstract Abstract: Users trust IoT apps to control and automate their smart devices. These apps necessarily have access to sensitive data to implement their functionality. However, users lack visibility into how their sensitive data is used, and often blindly trust the app developers. In this paper, we present IoTWATcH, a dynamic analysis tool that uncovers the privacy risks of IoT apps in real-time. We have designed and built IoTWATcH through a comprehensive IoT privacy survey addressing the privacy needs of users. IoTWATCH operates in four phases: (a) it provides users with an interface to specify their privacy preferences at app install time, (b) it adds extra logic to an app’s source code to collect both IoT data and their recipients at runtime, (c) it uses Natural Language Processing (NLP) techniques to construct a model that classifies IoT app data into intuitive privacy labels, and (d) it informs the users when their preferences do not match the privacy labels, exposing sensitive data leaks to users. We implemented and evaluated IoTWATcH on real IoT applications. Specifically, we analyzed 540 IoT apps to train the NLP model and evaluate its effectiveness. IoTWATcH yields an average 94.25% accuracy in classifying IoT app data into privacy labels with only 105 ms additional latency to an app’s execution.

Ensuring that information flowing through a network is secure from manipulation and eavesdropping by unauthorized parties is an important task for network administrators. Many cyber attacks rely on a lack of network-level information flow controls to successfully compromise a victim network. Once an adversary exploits an initial entry point, they can eavesdrop and move laterally within the network (e.g., scan and penetrate internal nodes) to further their malicious goals. In this article, we propose a novel multilevel security (MLS) framework to enforce a secure inter-node information flow policy within the network and therein vastly reduce the attack surface available to an adversary who has penetrated it. In contrast to prior work on multilevel security in computer networks which relied on enforcing the policy at network endpoints, we leverage the centralization of software-defined networks (SDNs) by moving the task to the controller and providing this service transparently to all network nodes. Our framework, MLSNet, formalizes the generation of a policy compliant network configuration (i.e., set of flow rules on the SDN switches) as network optimization problems, with the objectives of (1) maximizing the number of flows satisfying all security constraints and (2) minimizing the security cost of routing any remaining flows to guarantee availability. We demonstrate that MLSNet can securely and efficiently route flows that satisfy the security constraints and route the remaining flows with a minimal security cost (e.g., route >85% of flows, where the heuristic achieves 89% and 87% of the optimal solutions for the optimization problems).

The performance-driven design of SDN architectures leaves many security vulnerabilities, a notable one being the communication bottleneck between the controller and the switches. Functioning as a cache between the controller and the switches, the flow table mitigates this bottleneck by caching flow rules received from the controller at each switch, but is very limited in size due to the high cost and power consumption of the underlying storage medium. It thus presents an easy target for attacks. Observing that many existing defenses are based on simplistic attack models, we develop a model of intelligent attacks that exploit specific cache-like behaviors of the flow table to infer its internal configuration and state, and then design attack parameters accordingly. Our evaluations show that such attacks can accurately expose the internal parameters of the target flow table and cause measurable damage with the minimum effort.

The performance-driven design of SDN architectures leaves many security vulnerabilities, a notable one being the communication bottleneck between the controller and the switches. Functioning as a cache between the controller and the switches, the flow table mitigates this bottleneck by caching flow rules received from the controller at each switch, but is very limited in size due to the high cost and power consumption of the underlying storage medium. It thus presents an easy target for attacks. Observing that many existing defenses are based on simplistic attack models, we develop a model of intelligent attacks that exploit specific cache-like behaviors of the flow table to infer its internal configuration and state, and then design attack parameters accordingly. Our evaluations show that such attacks can accurately expose the internal parameters of the target flow table and cause measurable damage with the minimum effort.

In a smart home system, multiple users have access to multiple devices, typically through a dedicated app installed on a mobile device. Traditional access control mechanisms consider one unique trusted user that controls the access to the devices. However, multi-user multi-device smart home settings pose fundamentally different challenges to traditional single-user systems. For instance, in a multi-user environment, users have conflicting, complex, and dynamically changing demands on multiple devices, which cannot be handled by traditional access control techniques. To address these challenges, in this paper, we introduce Kratos, a novel multi-user and multi-device-aware access control mechanism that allows smart home users to flexibly specify their access control demands. Kratos has three main components: user interaction module, back-end server, and policy manager. Users can specify their desired access control settings using the interaction module which are translated into access control policies in the backend server. The policy manager analyzes these policies and initiates negotiation between users to resolve conflicting demands and generates final policies. We implemented Kratos and evaluated its performance on real smart home deployments featuring multi-user scenarios with a rich set of configurations (309 different policies including 213 demand conflicts and 24 restriction policies). These configurations included five different threats associated with access control mechanisms. Our extensive evaluations show that Kratos is very effective in resolving conflicting access control demands with minimal overhead, and robust against different attacks.

Data sharing among partners---users, companies, organizations---is crucial for the advancement of collaborative machine learning in many domains such as healthcare, finance, and security. Sharing through secure computation and other means allow these partners to perform privacy-preserving computations on their private data in controlled ways. However, in reality, there exist complex relationships among members (partners). Politics, regulations, interest, trust, data demands and needs prevent members from sharing their complete data. Thus, there is a need for a mechanism to meet these conflicting relationships on data sharing. This paper presents, an approach to exchange data among members who have complex relationships. A novel policy language, CPL, that allows members to define the specifications of data exchange requirements is introduced. With CPL, members can easily assert who and what to exchange through their local policies and negotiate a global sharing agreement. The agreement is implemented in a distributed privacy-preserving model that guarantees sharing among members will comply with the policy as negotiated. The use of Curie is validated through an example healthcare application built on recently introduced secure multi-party computation and differential privacy frameworks, and policy and performance trade-offs are explored.

VM migration is an effective countermeasure against attempts at malicious co-residency. In this paper, our overarching objectives are: (a) to get an in-depth understanding of the ways and effectiveness with which an attacker can launch attacks toward achieving co-residency and (b) to design migration policies that are very effective in thwarting malicious co-residency, but are thrifty in terms of the bandwidth and downtime costs that are incurred with live migration. Toward achieving our goals, we first undertake an experimental study on Amazon EC2 to obtain an in-depth understanding of the side-channels, through which an attacker can use to ascertain co-residency with a victim. Here, in this paper, we identify a new set of stealthy side-channel attacks which we show to be more effective than the currently available attacks toward verifying co-residency. We also build a simple model that can be used for estimating co-residency times based on very few measurements on a given cloud platform, to account for varying attacker capabilities. Based on the study, we develop a set of guidelines to determine under what conditions the victim VM migrations should be triggered, given the performance costs in terms of bandwidth and downtime, which a user is willing to bear. Through extensive experiments on our private in-house cloud, we show that the migrations, using our guidelines, can limit the fraction of the time that an attacker VM co-resides with a victim VM to about 1% of the time with the bandwidth costs of a few MB and downtimes of a few seconds per day per VM migrated.

For well over a quarter century, detection systems have been driven by models learned from input features collected from real or simulated environments. An artifact (e.g., network event, potential malware sample, suspicious email) is deemed malicious or non-malicious based on its similarity to the learned model at runtime. However, the training of the models has been historically limited to only those features available at runtime. In this paper, we consider an alternate learning approach that trains models using privileged information--features available at training time but not at runtime--to improve the accuracy and resilience of detection systems. In particular, we adapt and extend recent advances in knowledge transfer, model influence, and distillation to enable the use of forensic or other data unavailable at runtime in a range of security domains. An empirical evaluation shows that privileged information increases precision and recall over a system with no privileged information: we observe up to 7.7% relative decrease in detection error for fast-flux bot detection, 8.6% for malware traffic detection, 7.3% for malware classification, and 16.9% for face recognition. We explore the limitations and applications of different privileged information techniques in detection systems. Such techniques provide a new means for detection systems to learn from data that would otherwise not be available at runtime.

Today's IoT systems include event-driven smart applications (apps) that interact with sensors and actuators. A problem specific to IoT systems is that buggy apps, unforeseen bad app interactions, or device/communication failures, can cause unsafe and dangerous physical states. Detecting flaws that lead to such states, requires a holistic view of installed apps, component devices, their configurations, and more importantly, how they interact. In this paper, we design IotSan, a novel practical system that uses model checking as a building block to reveal "interaction-level" flaws by identifying events that can lead the system to unsafe states. In building IotSan, we design novel techniques tailored to IoT systems, to alleviate the state explosion associated with model checking. IotSan also automatically translates IoT apps into a format amenable to model checking. Finally, to understand the root cause of a detected vulnerability, we design an attribution mechanism to identify problematic and potentially malicious apps. We evaluate IotSan on the Samsung SmartThings platform. From 76 manually configured systems, IotSan detects 147 vulnerabilities. We also evaluate IotSan with malicious SmartThings apps from a previous effort. IotSan detects the potential safety violations and also effectively attributes these apps as malicious.

Advances in machine learning (ML) in recent years have enabled a dizzying array of applications such as data analytics, autonomous systems, and security diagnostics. ML is now pervasive-new systems and models are being deployed in every domain imaginable, leading to widespread deployment of software based inference and decision making. There is growing recognition that ML exposes new vulnerabilities in software systems, yet the technical community's understanding of the nature and extent of these vulnerabilities remains limited. We systematize findings on ML security and privacy, focusing on attacks identified on these systems and defenses crafted to date.We articulate a comprehensive threat model for ML, and categorize attacks and defenses within an adversarial framework. Key insights resulting from works both in the ML and security communities are identified and the effectiveness of approaches are related to structural elements of ML algorithms and the data used to train them. In particular, it is apparent that constructing a theoretical understanding of the sensitivity of modern ML algorithms to the data they analyze, à la PAC theory, will foster a science of security and privacy in ML.

Machine learning (ML) models, e.g., deep neural networks (DNNs), are vulnerable to adversarial examples: malicious inputs modified to yield erroneous model outputs, while appearing unmodified to human observers. Potential attacks include having malicious content like malware identified as legitimate or controlling vehicle behavior. Yet, all existing adversarial example attacks require knowledge of either the model internals or its training data. We introduce the first practical demonstration of an attacker controlling a remotely hosted DNN with no such knowledge. Indeed, the only capability of our black-box adversary is to observe labels given by the DNN to chosen inputs. Our attack strategy consists in training a local model to substitute for the target DNN, using inputs synthetically generated by an adversary and labeled by the target DNN. We use the local substitute to craft adversarial examples, and find that they are misclassified by the targeted DNN. To perform a real-world and properly-blinded evaluation, we attack a DNN hosted by MetaMind, an online deep learning API. We find that their DNN misclassifies 84.24% of the adversarial examples crafted with our substitute. We demonstrate the general applicability of our strategy to many ML techniques by conducting the same attack against models hosted by Amazon and Google, using logistic regression substitutes. They yield adversarial examples misclassified by Amazon and Google at rates of 96.19% and 88.94%. We also find that this black-box attack strategy is capable of evading defense strategies previously found to make adversarial example crafting harder.

Online social network (OSN) discussion groups are exerting significant effects on political dialogue. In the absence of access control mechanisms, any user can contribute to any OSN thread. Individuals can exploit this characteristic to execute targeted attacks, which increases the potential for subsequent malicious behaviors such as phishing and malware distribution. These kinds of actions will also disrupt bridges among the media, politicians, and their constituencies. For the concern of Security Management, blending malicious cyberattacks with online social interactions has introduced a brand new challenge. In this paper we describe our proposal for a novel approach to studying and understanding the strategies that attackers use to spread malicious URLs across Facebook discussion groups. We define and analyze problems tied to predicting the potential for attacks focused on threads created by news media organizations. We use a mix of macro static features and the micro dynamic evolution of posts and threads to identify likely targets with greater than 90% accuracy. One of our secondary goals is to make such predictions within a short (10 minute) time frame. It is our hope that the data and analyses presented in this paper will support a better understanding of attacker strategies and footprints, thereby developing new system management methodologies in handing cyber attacks on social networks.

Live virtual machine (VM) migration is commonly used for enabling dynamic resource or fault management, or for load balancing in datacenters or cloud platforms. A service hosted by a VM may also be migrated to prevent its visibility to an external adversary who may seek to disrupt its operation by launching a DDoS attack against it. We first show that current systems cannot adequately hide a VM migration from an external adversary. The key reason for this is that a migration typically manifests a traffic pattern with distinguishable statistical properties. We introduce two new attacks that can allow an adversary to effectively track a migration in progress, by leveraging observations of these properties. As our primary contribution, we design and implement a stealth migration framework that causes migration traffic to be indistinguishable from regular Internet traffic, with a negligible latency overhead of approximately 0.37 seconds, on average.

Attacker VMs try to co-reside with victim VMs on the same physical infrastructure as a precursor to launching attacks that target information leakage. VM migration is an effective countermeasure against attempts at malicious co-residency. In this paper, we first undertake an experimental study on Amazon EC2 to obtain an in-depth understanding of the side-channels an attacker can use to ascertain co-residency with a victim. Here, we identify a new set of stealthy side-channel attacks which, we show to be more effective than currently available attacks towards verifying co-residency. Based on the study, we develop a set of guidelines to determine under what conditions victim VM migrations should be triggered given performance costs in terms of bandwidth and downtime, that a user is willing to bear. Via extensive experiments on our private in-house cloud, we show that migrations using our guidelines can limit the fraction of the time that an attacker VM co-resides with a victim VM to about 1 % of the time with bandwidth costs of a few MB and downtimes of a few seconds, per day per VM migrated.

Application containers, such as those provided by Docker, have recently gained popularity as a solution for agile and seamless software deployment. These light-weight virtualization environments run applications that are packed together with their resources and configuration information, and thus can be deployed across various software platforms. Unfortunately, the ease with which containers can be created is oftentimes a double-edged sword, encouraging the packaging of logically distinct applications, and the inclusion of significant amount of unnecessary components, within a single container. These practices needlessly increase the container size-sometimes by orders of magnitude. They also decrease the overall security, as each included component-necessary or not-may bring in security issues of its own, and there is no isolation between multiple applications packaged within the same container image. We propose algorithms and a tool called Cimplifier, which address these concerns: given a container and simple user-defined constraints, our tool partitions it into simpler containers, which (i) are isolated from each other, only communicating as necessary, and (ii) only include enough resources to perform their functionality. Our evaluation on real-world containers demonstrates that Cimplifier preserves the original functionality, leads to reduction in image size of up to 95%, and processes even large containers in under thirty seconds.

The basis for all IPv4 network communication is the address resolution protocol (ARP), which maps an IP address to a device's media access control identifier. ARP has long been recognized as vulnerable to spoofing and other attacks, and past proposals to secure the protocol have often involved in modifying the basic protocol. Similarly, neighbor discovery protocol (NDP) is the basis for all IPv6 network communication, yet suffers from the same vulnerabilities as ARP. This paper introduces arpsec, a secure ARP/RARP protocol suite which does not require protocol modification, enables continual verification of the identity of the target (respondent) machine by introducing an address binding repository derived using a formal logic that bases additions to a host's ARP cache on a set of operational rules and properties, utilizes the trusted platform module (TPM), a commodity component now present in the vast majority of modern computers, to augment the logic-prover-derived assurance when needed, with TPM-facilitated attestations of system state achieved at viably low-processing cost, and supports IPv6 NDP (ndpsec) by extension of our previous work. Using commodity TPMs as our attestation base, we show that arpsec incurs an overhead ranging from 7% to 15.4% over the standard Linux ARP implementation, a comparable overhead against the standard Linux NDP implementation, and provides a first step towards a formally secure and trustworthy networking stack for both IPv4 and IPv6.

Advanced targeted cyber attacks often rely on reconnaissance missions to gather information about potential targets, their characteristics and location to identify vulnerabilities in a networked environment. Advanced network scanning techniques are often used for this purpose and are automatically executed by malware infected hosts. In this paper, we formally define network deception to defend reconnaissance and develop a reconnaissance deception system, which is based on software defined networking, to achieve deception by simulating virtual topologies. Our system thwarts network reconnaissance by delaying the scanning techniques of adversaries and invalidating their collected information, while limiting the performance impact on benign network traffic. By simulating the topological as well as physical characteristics of networks, we introduce a system which deceives malicious network discovery and reconnaissance techniques with virtual information, while limiting the information an attacker is able to harvest from the true underlying system. This approach shows a novel defense technique against adversarial reconnaissance missions which are required for targeted cyber attacks such as advanced persistent threats in highly connected environments. The defense steps of our system aim to invalidate an attackers information, delay the process of finding vulnerable hosts and identify the source of adversarial reconnaissance within a network.

Privacy is an important aspect of network communications, but privacy protocols require an investment of network resources. For any such protocol to be of use, we need to understand quantitatively how much privacy to expect, as well as the tradeoff between privacy and other network properties, for any given configuration of networks and parameters. We develop a practical privacy measure and protocol model for multichannel secret sharing protocols which integrates privacy and measurable network properties, deriving optimality results for the overall privacy and performance of these protocols. After proving these results, we evaluate the effectiveness of our model by providing a reference implementation and comparing its behavior to the optimality results derived from the model. In our benchmarks, the behavior of this proof-of-concept protocol matched that which is predicted by our model, furthermore, our results demonstrate the feasibility of implementing secret sharing protocols which transmit at a rate within 3-4% of optimal. This model and its results allow us to understand quantitatively the tradeoffs between privacy and network performance in secret-sharing based protocols.

Deep learning takes advantage of large datasets and computationally efficient training algorithms to outperform other approaches at various machine learning tasks. However, imperfections in the training phase of deep neural networks make them vulnerable to adversarial samples: inputs crafted by adversaries with the intent of causing deep neural networks to misclassify. In this work, we formalize the space of adversaries against deep neural networks (DNNs) and introduce a novel class of algorithms to craft adversarial samples based on a precise understanding of the mapping between inputs and outputs of DNNs. In an application to computer vision, we show that our algorithms can reliably produce samples correctly classified by human subjects but misclassified in specific targets by a DNN with a 97% adversarial success rate while only modifying on average 4.02% of the input features per sample. We then evaluate the vulnerability of different sample classes to adversarial perturbations by defining a hardness measure. Finally, we describe preliminary work outlining defenses against adversarial samples by defining a predictive measure of distance between a benign input and a target classification.

Static analysis has been successfully used in many areas, from verifying mission-critical software to malware detection. Unfortunately, static analysis often produces false positives, which require significant manual effort to resolve. In this paper, we show how to overlay a probabilistic model, trained using domain knowledge, on top of static analysis results, in order to triage static analysis results. We apply this idea to analyzing mobile applications. Android application components can communicate with each other, both within single applications and between different applications. Unfortunately, techniques to statically infer Inter-Component Communication (ICC) yield many potential inter-component and inter-application links, most of which are false positives. At large scales, scrutinizing all potential links is simply not feasible. We therefore overlay a probabilistic model of ICC on top of static analysis results. Since computing the inter-component links is a prerequisite to inter-component analysis, we introduce a formalism for inferring ICC links based on set constraints. We design an efficient algorithm for performing link resolution. We compute all potential links in a corpus of 11,267 applications in 30 minutes and triage them using our probabilistic approach. We find that over 95.1% of all 636 million potential links are associated with probability values below 0.01 and are thus likely unfeasible links. Thus, it is possible to consider only a small subset of all links without significant loss of information. This work is the first significant step in making static inter-application analysis more tractable, even at large scales.

Android security and privacy research has boomed in recent years, far outstripping investigations of other appified platforms. However, despite this attention, research efforts are fragmented and lack any coherent evaluation framework. We present a systematization of Android security and privacy research with a focus on the appification of software systems. To put Android security and privacy research into context, we compare the concept of appification with conventional operating system and software ecosystems. While appification has improved some issues (e.g., market access and usability), it has also introduced a whole range of new problems and aggravated some problems of the old ecosystems (e.g., coarse and unclear policy, poor software development practices). Some of our key findings are that contemporary research frequently stays on the beaten path instead of following unconventional and often promising new routes. Many security and privacy proposals focus entirely on the Android OS and do not take advantage of the unique features and actors of an appified ecosystem, which could be used to roll out new security mechanisms less disruptively. Our work highlights areas that have received the larger shares of attention, which attacker models were addressed, who is the target, and who has the capabilities and incentives to implement the countermeasures. We conclude with lessons learned from comparing the appified with the old world, shedding light on missed opportunities and proposing directions for future research.

Any individual that re-registers an expired domain implicitly inherits the residual trust associated with the domain's prior use. We find that adversaries can, and do, use malicious re-registration to exploit domain ownership changes - undermining the security of both users and systems. In fact, we find that many seemingly disparate security problems share a root cause in residual domain trust abuse. With this study we shed light on the seemingly unnoticed problem of residual domain trust by measuring the scope and growth of this abuse over the past six years. During this time, we identified 27,758 domains from public blacklists and 238,279 domains resolved by malware that expired and then were maliciously re-registered. To help address this problem, we propose a technical remedy and discuss several policy remedies. For the former, we develop Alembic, a lightweight algorithm that uses only passive observations from the Domain Name System (DNS) to flag potential domain ownership changes. We identify several instances of residual trust abuse using this algorithm, including an expired APT domain that could be used to revive existing infections.

Deep learning algorithms have been shown to perform extremely well on many classical machine learning problems. However, recent studies have shown that deep learning, like other machine learning techniques, is vulnerable to adversarial samples: inputs crafted to force a deep neural network (DNN) to provide adversary-selected outputs. Such attacks can seriously undermine the security of the system supported by the DNN, sometimes with devastating consequences. For example, autonomous vehicles can be crashed, illicit or illegal content can bypass content filters, or biometric authentication systems can be manipulated to allow improper access. In this work, we introduce a defensive mechanism called defensive distillation to reduce the effectiveness of adversarial samples on DNNs. We analytically investigate the generalizability and robustness properties granted by the use of defensive distillation when training DNNs. We also empirically study the effectiveness of our defense mechanisms on two DNNs placed in adversarial settings. The study shows that defensive distillation can reduce effectiveness of sample creation from 95% to less than 0.5% on a studied DNN. Such dramatic gains can be explained by the fact that distillation leads gradients used in adversarial sample creation to be reduced by a factor of 1030. We also find that distillation increases the average minimum number of features that need to be modified to create adversarial samples by about 800% on one of the DNNs we tested.

Many program analyses require statically inferring the possible values of composite types. However, current approaches either do not account for correlations between object fields or do so in an ad hoc manner. In this paper, we introduce the problem of composite constant propagation. We develop the first generic solver that infers all possible values of complex objects in an interprocedural, flow and context-sensitive manner, taking field correlations into account. Composite constant propagation problems are specified using COAL, a declarative language. We apply our COAL solver to the problem of inferring Android Inter-Component Communication (ICC) values, which is required to understand how the components of Android applications interact. Using COAL, we model ICC objects in Android more thoroughly than the state-of-the-art. We compute ICC values for 489 applications from the Google Play store. The ICC values we infer are substantially more precise than previous work. The analysis is efficient, taking two minutes per application on average. While this work can be used as the basis for many whole-program analyses of Android applications, the COAL solver can also be used to infer the values of composite objects in many other contexts.

The basis for all IPv4 network communication is the Address Resolution Protocol (ARP), which maps an IP address to a device's Media Access Control (MAC) identifier. ARP has long been recognized as vulnerable to spoofing and other attacks, and past proposals to secure the protocol have often involved modifying the basic protocol.

Flaws in cryptosystem implementations, such as the Heartbleed bug, render common confidentiality mechanisms ineffective. Defending in depth when this happens would require a different means of providing confidentiality, which could then be layered with existing cryptosystems. This paper presents MICSS, a network protocol which uses multichannel secret sharing rather than encryption to protect data confidentiality. The MICSS protocol ensures perfect secrecy against an (n-1)-channel attacker and operates at line speed in a three-channel throughput benchmark. MICSS provides a practical means of securing network communications, and it layers seamlessly with cryptosystems to mitigate the effects of implementation flaws.

Shake Them All is a popular "Wallpaper" application exceeding millions of downloads on Google Play. At installation, this application is given permission to (1) access the Internet (for updating wallpapers) and (2) use the device microphone (to change background following noise changes). With these permissions, the application could silently record user conversations and upload them remotely. To give more confidence about how Shake Them All actually processes what it records, it is necessary to build a precise analysis tool that tracks the flow of any sensitive data from its source point to any sink, especially if those are in different components. Since Android applications may leak private data carelessly or maliciously, we propose IccTA, a static taint analyzer to detect privacy leaks among components in Android applications. IccTA goes beyond state-of-the-art approaches by supporting inter- component detection. By propagating context information among components, IccTA improves the precision of the analysis. IccTA outperforms existing tools on two benchmarks for ICC-leak detectors: DroidBench and ICC-Bench. Moreover, our approach detects 534 ICC leaks in 108 apps from MalGenome and 2,395 ICC leaks in 337 apps in a set of 15,000 Google Play apps.

Many program analyses require statically inferring the possible values of composite types. However, current approaches either do not account for correlations between object fields or do so in an ad hoc manner. In this paper, we introduce the problem of composite constant propagation. We develop the first generic solver that infers all possible values of complex objects in an interprocedural, flow and context-sensitive manner, taking field correlations into account. Composite constant propagation problems are specified using COAL, a declarative language. We apply our COAL solver to the problem of inferring Android Inter-Component Communication (ICC) values, which is required to understand how the components of Android applications interact. Using COAL, we model ICC objects in Android more thoroughly than the state-of-the-art. We compute ICC values for 460 applications from the Play store. The ICC values we infer are substantially more precise than previous work. The analysis is efficient, taking slightly over two minutes per application on average. While this work can be used as the basis for many whole-program analyses of Android applications, the COAL solver can also be used to infer the values of composite objects in many other contexts.

In 2011, Adblock Plus---the most widely-used ad blocking software---began to permit some advertisements as part of their Acceptable Ads program. Under this program, some ad networks and content providers pay to have their advertisements shown to users. Such practices have been controversial among both users and publishers. In a step towards informing the discussion about these practices, we present the first comprehensive study of the Acceptable Ads program. Specifically, we characterize which advertisements are allowed and how the whitelisting has changed since its introduction in 2011. We show that the list of filters used to whitelist acceptable advertisements has been updated on average every 1.5 days and grew from 9 filters in 2011 to over 5,900 in the Spring of 2015. More broadly, the current whitelist triggers filters on 59% of the top 5,000 websites. Our measurements also show that the program allows advertisements on 2.6 million parked domains. Lastly, we take the lessons learned from our analysis and suggest ways to improve the transparency of the whitelisting process.

Today's smartphones are a ubiquitous source of private and confidential data. At the same time, smartphone users are plagued by carelessly programmed apps that leak important data by accident, and by malicious apps that exploit their given privileges to copy such data intentionally. While existing static taint-analysis approaches have the potential of detecting such data leaks ahead of time, all approaches for Android use a number of coarse-grain approximations that can yield high numbers of missed leaks and false alarms.

In recent years, the Android operating system has had an explosive growth in the number of applications containing third-party libraries for different purposes. In this paper, we identify three library-centric threats in the real-world Android application markets: (i) the library modification threat, (ii) the masquerading threat and (iii) the aggressive library threat. These three threats cannot effectively be fully addressed by existing defense mechanisms such as software analysis, anti-virus software and anti-repackaging techniques. To mitigate these threats, we propose Duet, a library integrity verification tool for Android applications at application stores. This is non-trivial because the Android application build process merges library code and application-specific logic into a single binary file. Our approach uses reverse-engineering to achieve integrity verification. We implemented a full working prototype of Duet. In a dataset with 100,000 Android applications downloaded from Google Play between February 2012 and September 2013, we verify integrity of 15 libraries. On average, 80.50% of libraries can pass the integrity verification. In-depth analysis indicates that code insertion, obfuscation, and optimization on libraries by application developers are the primary reasons for not passing integrity verification. The evaluation results not only indicate that Duet is an effective tool to mitigate library-centric attacks, but also provide empirical insight into the library integrity situation in the wild.

Today’s smartphone operating systems frequently fail to provide users with visibility into how third-party applications collect and share their private data. We address these shortcomings with TaintDroid, an efficient, system-wide dynamic taint tracking and analysis system capable of simultaneously tracking multiple sources of sensitive data. TaintDroid enables realtime analysis by leveraging Android’s virtualized execution environment. TaintDroid incurs only 32% performance overhead on a CPU-bound microbenchmark and imposes negligible overhead on interactive third-party applications. Using TaintDroid to monitor the behavior of 30 popular third-party Android applications, in our 2010 study we found 20 applications potentially misused users’ private information; so did a similar fraction of the tested applications in our 2012 study. Monitoring the flow of privacy-sensitive data with TaintDroid provides valuable input for smartphone users and security service firms seeking to identify misbehaving applications.

The articles in this special section focus on trust, network security, and privacy deployed in parallel and distributed systems.

Data provenance---a record of the origin and evolution of data in a system---is a useful tool for forensic analysis. However, existing provenance collection mechanisms fail to achieve sufficient breadth or fidelity to provide a holistic view of a system's operation over time. We present Hi-Fi, a kernel-level provenance system which leverages the Linux Security Modules framework to collect high-fidelity whole-system provenance. We demonstrate that Hi-Fi is able to record a variety of malicious behavior within a compromised system. In addition, our benchmarks show the collection overhead from Hi-Fi to be less than 1% for most system calls and 3% in a representative workload, while simultaneously generating a system measurement that fully reflects system evolution. In this way, we show that we can collect broad, high-fidelity provenance data which is capable of supporting detailed forensic analysis.

Programmable Logic Controllers (PLCs) drive the behavior of industrial control systems according to uploaded programs. It is now known that PLCs are vulnerable to the uploading of malicious code that can have severe physical consequences. What is not understood is whether an adversary with no knowledge of the PLC's interface to the control system can execute a damaging, targeted, or stealthy attack against a control system using the PLC. In this paper, we present SABOT, a tool that automatically maps the control instructions in a PLC to an adversary-provided specification of the target control system's behavior. This mapping recovers sufficient semantics of the PLC's internal layout to instantiate arbitrary malicious controller code. This lowers the prerequisite knowledge needed to tailor an attack to a control system. SABOT uses an incremental model checking algorithm to map a few plant devices at a time, until a mapping is found for all adversary-specified devices. At this point, a malicious payload can be compiled and uploaded to the PLC. Our evaluation shows that SABOT correctly compiles payloads for all tested control systems when the adversary correctly specifies full system behavior, and for 4 out of 5 systems in most cases where there where unspecified features. Furthermore, SABOT completed all analyses in under 2 minutes.

Smart electric meters pose a substantial threat to the privacy of individuals in their own homes. Combined with non-intrusive load monitors, smart meter data can reveal precise home appliance usage information. An emerging solution to behavior leakage in smart meter measurement data is the use of battery-based load hiding. In this approach, a battery is used to store and supply power to home devices at strategic times to hide appliance loads from smart meters. A few such battery control algorithms have already been studied in the literature, but none have been evaluated from an adversarial point of view. In this paper, we first consider two well known battery privacy algorithms, Best Effort (BE) and Non-Intrusive Load Leveling (NILL), and demonstrate attacks that recover precise load change information, which can be used to recover appliance behavior information, under both algorithms. We then introduce a stepping approach to battery privacy algorithms that fundamentally differs from previous approaches by maximizing the error between the load demanded by a home and the external load seen by a smart meter. By design, precise load change recovery attacks are impossible. We also propose mutual-information based measurements to evaluate the privacy of different algorithms. We implement and evaluate four novel algorithms using the stepping approach, and show that under the mutual-information metrics they outperform BE and NILL.

The Android OS has emerged as the leading platform for SmartPhone applications. However, because Android applications are compiled from Java source into platform-specific Dalvik bytecode, existing program analysis tools cannot be used to evaluate their behavior. This paper develops and evaluates algorithms for retargeting Android applications received from markets to Java class files. The resulting Dare tool uses a new intermediate representation to enable fast and accurate retargeting. Dare further applies strong constraint solving to infer typing information and translates the 257 DVM opcodes using only 9 translation rules. It also handles cases where the input Dalvik bytecode is unverifiable. We evaluate Dare on 1,100 of the top applications found in the free section of the Android market and successfully retarget 99.99% of the 262,110 associated classes. Further, whereas existing tools can only fully retarget about half of these applications, Dare can recover over 99% of them. In this way, we open the door to users, developers and markets to use the vast array of program analysis tools to ensure the correct operation of Android applications.

ABSTRACT Smartphones are now ubiquitous. However, the security requirements of these relatively new systems and the applications they support are still being understood. As a result, the security infrastructure available in current smartphone operating systems is largely underdeveloped. In this paper, we consider the security requirements of smartphone applications and augment the existing Android operating system with a framework to meet them. We present Secure Application INTeraction (Saint), a modified infrastructure that governs install‐time permission assignment and their run‐time use as dictated by application provider policy. An in‐depth description of the semantics of application policy is presented. The architecture and technical detail of Saint are given, and areas for extension, optimization, and improvement are explored. We demonstrate through a concrete example and study of real‐world applications that Saint provides necessary utility for applications to assert and control the security decisions on the platform. Copyright © 2011 John Wiley & Sons, Ltd.

The web is a primary means of information sharing for most organizations and people. Currently, a recipient of web content knows nothing about the environment in which that information was generated other than the specific server from whence it came (and even that information can be unreliable). In this paper, we develop and evaluate the Spork system that uses the Trusted Platform Module (TPM) to tie the web server integrity state to the web content delivered to browsers, thus allowing a client to verify that the origin of the content was functioning properly when the received content was generated and/or delivered. We discuss the design and implementation of the Spork service and its browser-side Firefox validation extension. In particular, we explore the challenges and solutions of scaling the delivery of mixed static and dynamic content to a large number of clients using exceptionally slow TPM hardware. We perform an in-depth empirical analysis of the Spork system within Apache web servers. This analysis shows Spork can deliver nearly 8,000 static or over 6,500 dynamic integrity-measured web objects per second. More broadly, we identify how TPM-based content web services can scale to large client loads with manageable overheads and deliver integrity-measured content with manageable overhead.

The smart grid introduces concerns for the loss of consumer privacy; recently deployed smart meters retain and distribute highly accurate profiles of home energy use. These profiles can be mined by Non Intrusive Load Monitors (NILMs) to expose much of the human activity within the served site. This paper introduces a new class of algorithms and systems, called Non Intrusive Load Leveling (NILL) to combat potential invasions of privacy. NILL uses an in-residence battery to mask variance in load on the grid, thus eliminating exposure of the appliance-driven information used to compromise consumer privacy. We use real residential energy use profiles to drive four simulated deployments of NILL. The simulations show that NILL exposes only 1.1 to 5.9 useful energy events per day hidden amongst hundreds or thousands of similar battery-suppressed events. Thus, the energy profiles exhibited by NILL are largely useless for current NILM algorithms. Surprisingly, such privacy gains can be achieved using battery systems whose storage capacity is far lower than the residence's aggregate load average. We conclude by discussing how the costs of NILL can be offset by energy savings under tiered energy schedules.

Abstract Mobile phones have evolved from simple voice terminals into highly‐capable, general‐purpose computing platforms. While people are becoming increasingly more dependent on such devices to perform sensitive operations, protect secret data, and be available for emergency use, it is clear that phone operating systems are not ready to become mission‐critical systems. Through a pair of vulnerabilities and a simulated attack on a cellular network, we demonstrate that there are a myriad of unmanaged mechanisms on mobile phones, and that control of these mechanisms is vital to achieving reliable use. Through such vectors, mobile phones introduce a variety of new threats to their own applications and the telecommunications infrastructure itself. In this paper, we examine the requirements for providing effective mediation and access control for mobile phones. We then discuss the convergence of cellular networks with the Internet and its impact on effective resource management and quality of service. Based on these results, we argue for user devices that enable predictable behavior in a network—where their trusted computing bases can protect key applications and create predictable network impact. Copyright © 2010 John Wiley & Sons, Ltd.

Portable storage devices, such as key-chain USB devices, are ubiquitous. These devices are often used with impunity, with users repeatedly using the same storage device in open computer laboratories, Internet cafes, and on office and home computers. Consequently, they are the target of malware that exploit the data present or use them as a means to propagate malicious software. This paper presents the Kells mobile storage system. Kells limits untrusted or unknown systems from accessing sensitive data by continuously validating the accessing host's integrity state. We explore the design and operation of Kells, and implement a proof-of-concept USB 2.0 storage device on experimental hardware. Our analysis of Kells is twofold. We first prove the security of device operation (within a freshness security parameter Δt) using the LS2 logic of secure systems. Second, we empirically evaluate the performance of Kells. These experiments indicate nominal overheads associated with host validation, showing a worst case throughput overhead of 1.22% for read operations and 2.78% for writes.

The advanced metering infrastructure (AMI) is revolutionizing electrical grids. Intelligent AMI "smart meters" report real time usage data that enables efficient energy generation and use. However, aggressive deployments are outpacing security efforts: new devices from a dizzying array of vendors are being introduced into grids with little or no understanding of the security problems they represent. In this paper we develop an archetypal attack tree approach to guide penetration testing across multiple-vendor implementations of a technology class. In this, we graft archetypal attack trees modeling broad adversary goals and attack vectors to vendor-specific concrete attack trees. Evaluators then use the grafted trees as a roadmap to penetration testing. We apply this approach within AMI to model attacker goals such as energy fraud and denial of service. Our experiments with multiple vendors generate real attack scenarios using vulnerabilities identified during directed penetration testing, e.g., manipulation of energy usage data, spoofing meters, and extracting sensitive data from internal registers. More broadly, we show how we can reuse efforts in penetration testing to efficiently evaluate the increasingly large body of AMI technologies being deployed in the field.

The penetration of cellular networks worldwide and emergence of smart phones has led to a revolution in mobile content. Users consume diverse content when, for example, exchanging photos, playing games, browsing websites, and viewing multimedia. Current phone platforms provide protections for user privacy, the cellular radio, and the integrity of the OS itself. However, few offer protections to protect the content once it enters the phone. For example, MP3-based MMS or photo content placed on Android smart phones can be extracted and shared with impunity. In this paper, we explore the requirements and enforcement of digital rights management (DRM) policy on smart phones. An analysis of the Android market shows that DRM services should ensure: a) protected content is accessible only by authorized phones b) content is only accessible by provider-endorsed applications, and c) access is regulated by contextual constraints, e.g., used for a limited time, a maximum number of viewings, etc. The Porscha system developed in this work places content proxies and reference monitors within the Android middleware to enforce DRM policies embedded in received content. A pilot study controlling content obtained over SMS, MMS, and email illustrates the expressibility and enforcement of Porscha policies. Our experiments demonstrate that Porscha is expressive enough to articulate needed DRM policies and that their enforcement has limited impact on performance.

Portable storage devices, such as key-chain USB devices, are ubiquitous and used everywhere; users repeatedly use the same storage device in open computer laboratories, Internet cafes, and on office and home computers. Consequently, they are the target of malware that exploit the data present or use them as a means to propagate malicious software., e.g., Conficker and Agent.bz. We present the Kells mobile storage system, which limits untrusted or unknown systems from accessing sensitive data by continuously validating the accessing host's integrity state. We explore the design and operation of Kells, and implement a proof-of-concept USB 2.0 storage device of experimental hardware. Our experiments indicate nominal overheads associated with host validation, with a worst-case throughput overhead of 1.22% for reads and 2.78% for writes.

Proof of a user's identity is not always a sufficient means for making an authorization decision. In an increasing set of circumstances, knowledge of physical location provides additional and necessary context for making decisions about resource access. For example, sensitive information stored on a laptop (e.g. customer records, social security numbers, etc), may require additional protections if a user operates outside of an approved area. However, current localization techniques based on signal strength reporting or specialized hardware fail to achieve this goal. In this paper, we design, develop, deploy and measure a system which securely determines the location of a user to within one meter through using only off-the-shelf 802.11 and Bluetooth equipment. We apply this equipment in a two-phased challenge- response protocol: first determining the general area of the client in the Regionalization phase and then pinpointing it in the Localization phase. Using nearly 32,000 data points collected over 75 days, we argue that the stability of wireless networks over time creates easily distinguishable location profiles by which a client can be positioned. Additionally, we demonstrate the inherent ability of a two-phased protocol to discern a client's location information at a level of granularity no finer than is necessitated by policy. After discussing a number of applications, we build a location-based access control framework that automatically protects a white-listed set of resources through encryption when the user leaves specified areas. Our analyses show that this system provides a realistic and efficient means of incorporating unforgeable location information at the appropriate level of granularity into many authorization decisions.

Storage is increasingly becoming a vector for data compromise. Solutions for protecting on-disk data confidentiality and integrity to date have been limited in their effectiveness. Providing authenticated encryption, or simultaneous encryption with integrity information, is important to protect data at rest. In this paper, we propose that disks augmented with non-volatile storage (e.g., hybrid hard disks) and cryptographic processors (e.g., FDE drives) may provide a solution for authenticated encryption, storing security metadata within the drive itself to eliminate dependences on other parts of the system. We augment the DiskSim simulator with a flash simulator to evaluate the costs associated with managing operational overheads. These experiments show that proper tuning of system parameters can eliminate many of the costs associated with managing security metadata, with less than a 2% decrease in IOPS versus regular disks.

The web is now being used as a general platform for hosting distributed applications like wikis, bulletin board messaging systems and collaborative editing environments. Data from multiple applications originating at multiple sources all intermix in a single web browser, making sensitive data stored in the browser subject to a broad milieu of attacks (cross-site scripting, cross-site request forgery and others). The fundamental problem is that existing web infrastructure provides no means for enforcing end-to-end security on data. To solve this we design an architecture using mandatory access control (MAC) enforcement. We overcome the limitations of traditional MAC systems, implemented solely at the operating system layer, by unifying MAC enforcement across virtual machine, operating system, networking and application layers. We implement our architecture using Xen virtual machine management, SELinux at the operating system layer, labeled IPsec for networking and our own label-enforcing web browser, called FlowwolF. We tested our implementation and find that it performs well, supporting data intermixing while still providing end-to-end security guarantees.

Attributes define, classify, or annotate the datum to which they are assigned. However, traditional attribute architectures and cryptosystems are ill-equipped to provide security in the face of diverse access requirements and environments. In this paper, we introduce a novel secure information management architecture based on emerging attribute-based encryption (ABE) primitives. A policy system that meets the needs of complex policies is defined and illustrated. Based on the needs of those policies, we propose cryptographic optimizations that vastly improve enforcement efficiency. We further explore the use of such policies in two proposed applications: a HIPAA compliant distributed file system and a social network. A performance analysis and characterization of ABE primitives demonstrates the ability to reduce cryptographic costs by as much as 98% over previously proposed constructions. Through this, we demonstrate that our attribute system is an efficient solution for securely managing information in large, loosely-coupled, distributed systems.

As the Internet'sde factointerdomain routing protocol, the Border Gateway Protocol (BGP) is the glue that holds the disparate parts of the Internet together. A major limitation of BGP is its failure to adequately address security. Recent high-profile outages and security analyses clearly indicate that the Internet routing infrastructure is highly vulnerable. Moreover, the design of BGP and the ubiquity of its deployment have frustrated past efforts at securing interdomain routing. This paper considers the current vulnerabilities of the interdomain routing system and surveys both research and standardization efforts relating to BGP security. We explore the limitations and advantages of proposed security extensions to BGP, and explain why no solution has yet struck an adequate balance between comprehensive security and deployment cost.

Abstract Densely populated areas are increasingly filled with vulnerable wireless routers set up by unsophisticated users. In isolation, such routers appear to represent only a minor threat, but in aggregate, the threat can be much greater. We introduce the notion of malnets: networks of adversary‐controlled wireless routers targeted to a physical geography. Similar to Internet worms such as Slammer and Code‐Red, malnets are created by the recursive compromise of targeted devices. However, unlike their traditionally wired counterparts, malnet worms exploit only other routers that are within their transmission range. The malnet thus creates a parallel wireless infrastructure that is (a) completely under control of the adversary, and (b) spans a targeted physical area, creating a valuable infrastructure for a variety of virtual and physical attacks. We initially study the propagation characteristics of commercial routers and model inter‐router connectivity using publicly available war‐driving data. The resulting characterization is applied to well‐known epidemiological models to explore the success rates and speeds of malnet creation across cities such as New York, Atlanta, and Los Angles. Finally, we use a sampling of available exploits to demonstrate the construction of multi‐vector, multi‐platform worms capable of targeting wireless routers. Our analysis show that an adversary can potentially deploy a malnet of over 24,000 routers in Manhattan in less than 2,h. Through this work we show that malnets are not only feasible but can be efficiently deployed. Copyright © 2009 John Wiley & Sons, Ltd.

The SELinux mandatory access control (MAC) policy has recently added a multilevel security (MLS) model which is able to express a fine granularity of control over a subject's access rights. The problem is that the richness of the SELinux MLS model makes it impractical to manually evaluate that a given policy meets certain specific properties. To address this issue, we have modeled the SELinux MLS model, using a logical specification and implemented that specification in the Prolog language. Furthermore, we have developed some analyses for testing information flow properties of a given policy as well as an algorithm to determine whether one policy is compliant with another. We have implemented these analyses in Prolog and compiled our implementation into a tool for SELinux MLS policy analysis, called PALMS. Using PALMS, we verified some important properties of the SELinux MLS reference policy, namely that it satisfies the simple security condition and ⋆-property defined by Bell and LaPadula. We also evaluated whether the policy associated to a given application is compliant with the policy of the SELinux system in which it would be deployed.

The Web is a primary means of information sharing for most organizations and people. Currently, a recipient of Web content knows nothing about the environment in which that information was generated other than the specific server from whence it came (and even that information can be unreliable). In this paper, we develop and evaluate the Spork system that uses the trusted platform module (TPM) to tie the Web server integrity state to the Web content delivered to browsers, thus allowing a client to verify that the origin of the content was functioning properly when the received content was generated and/or delivered. We discuss the design and implementation of the Spork service and its browser-side Firefox validation extension. In particular, we explore the challenges and solutions of scaling the delivery of mixed static and dynamic content using exceptionally slow TPM hardware. We perform an in-depth empirical analysis of the Spork system within Apache Web servers. This analysis shows Spork can deliver nearly 8,000 static or over 7,000 dynamic integrity-measured Web objects per-second. More broadly, we identify how TPM-based content Web services can scale with manageable overheads and deliver integrity-measured content with manageable overhead.

Smartphones are now ubiquitous. However, the security requirements of these relatively new systems and the applications they support are still being understood. As a result, the security infrastructure available in current smartphone operating systems is largely underdeveloped. In this paper, we consider the security requirements of smartphone applications and augment the existing Android operating system with a framework to meet them. We present Secure Application INTeraction (Saint), a modified infrastructure that governs install-time permission assignment and their run-time use as dictated by application provider policy. An in-depth description of the semantics of application policy is presented. The architecture and technical detail of Saint is given, and areas for extension, optimization, and improvement explored. As we show through concrete example, Saint provides necessary utility for applications to assert and control the security decisions on the platform.

Emerging distributed computing architectures, such as grid and cloud computing, depend on the high integrity execution of each system in the computation. While integrity measurement enables systems to generate proofs of their integrity to remote parties, we find that current integrity measurement approaches are insufficient to prove runtime integrity for systems in these architectures. Integrity measurement approaches that are flexible enough have an incomplete view of runtime integrity, possibly leading to false integrity claims, and approaches that provide comprehensive integrity do so only for computing environments that are too restrictive. In this paper, we propose an architecture for building comprehensive runtime integrity proofs for general purpose systems in distributed computing architectures. In this architecture, we strive for classical integrity, using an approximation of the Clark-Wilson integrity model as our target. Key to building such integrity proofs is a carefully crafted host system whose long-term integrity can be justified easily using current techniques and a new component, called a VM verifier, which comprehensively enforces our integrity target on VMs. We have built a prototype based on the Xen virtual machine system for SELinux VMs, and find that distributed compilation can be implemented, providing accurate proofs of our integrity target with less than 4% overhead.

Users have begun downloading an increasingly large number of mobile phone applications in response to advancements in handsets and wireless networks. The increased number of applications results in a greater chance of installing Trojans and similar malware. In this paper, we propose the Kirin security service for Android, which performs lightweight certification of applications to mitigate malware at install time. Kirin certification uses security rules, which are templates designed to conservatively match undesirable properties in security configuration bundled with applications. We use a variant of security requirements engineering techniques to perform an in-depth security analysis of Android to produce a set of rules that match malware characteristics. In a sample of 311 of the most popular applications downloaded from the official Android Market, Kirin and our rules found 5 applications that implement dangerous functionality and therefore should be installed with extreme caution. Upon close inspection, another five applications asserted dangerous rights, but were within the scope of reasonable functional needs. These results indicate that security configuration bundled with Android applications provides practical means of detecting malware.

The vast expansion of interconnectivity with the Internet and the rapid evolution of highly-capable but largely insecure mobile devices threatens cellular networks. In this paper, we characterize the impact of the large scale compromise and coordination of mobile phones in attacks against the core of these networks. Through a combination of measurement, simulation and analysis, we demonstrate the ability of a botnet composed of as few as 11,750 compromised mobile phones to degrade service to area-code sized regions by 93%. As such attacks are accomplished through the execution of network service requests and not a constant stream of phone calls, users are unlikely to be aware of their occurrence. We then investigate a number of significant network bottlenecks, their impact on the density of compromised nodes per base station and how they can be avoided. We conclude by discussing a number of countermeasures that may help to partially mitigate the threats posed by such attacks.

The development and maintenance of network device configurations is one of the central challenges faced by large network providers. Current network management systems fail to meet this challenge primarily because of their inability to adapt to rapidly evolving customer and provider-network needs, and because of mismatches between the conceptual models of the tools and the services they must support. In this paper, we present the Presto configuration management system that attempts to address these failings in a comprehensive and flexible way. Developed for and used during the last 5 years within a large ISP network, Presto constructs device-native configurations based on the composition of configlets representing different services or service options. Configlets are compiled by extracting and manipulating data from external systems as directed by the Presto configuration scripting and template language. We outline the configuration management needs of large-scale network providers, introduce the PRESTO system and configuration language, and reflect upon our experiences developing PRESTO configured VPN and VoIP services. In doing so, we describe how PRESTO promotes healthy configuration management practices.

Structured peer-to-peer (P2P) systems have grown enormously because of their scalability, efficiency, and reliability. These systems assign a unique identifier to each user and object. However, current assignment schemes allow an adversary to carefully select user IDs and/or simultaneously obtain many pseudo-identities-ultimately leading to an ability to disrupt the P2P system in very targeted and dangerous ways. In this paper, we propose novel ID assignment protocols based on identity-based cryptography. This approach permits the acquisition of node IDs to be tightly regulated without many of the complexities and costs associated with traditional certificate solutions. We broadly consider the security requirements of ID assignment and present three protocols representing distinct threat and trust models. A detailed empirical study of the protocols is given. Our analysis shows that the cost of our identity-based protocols is nominal, and that the associated identity services can scale to millions of users using a limited number of servers.

Main memory contains transient information for all resident applications. However, if memory chip contents survives power-off, e.g., via freezing DRAM chips, sensitive data such as passwords and keys can be extracted. Main memory persistence will soon be the norm as recent advancements in MRAM and FeRAM position non-volatile memory technologies for widespread deployment in laptop, desktop, and embedded system main memory. Unfortunately, the same properties that provide energy efficiency, tolerance against power failure, and "instant-on'' power-up also subject systems to offline memory scanning. In this paper, we propose a memory encryption control unit (MECU) that provides memory confidentiality during system suspend and across reboots. The MECU encrypts all memory transfers between the processor-local level 2 cache and main memory to ensure plaintext data is never written to the persistent medium. The MECU design is outlined and performance and security trade-offs considered. We evaluate a MECU-enhanced architecture using the SimpleScalar hardware simulation framework on several hardware benchmarks. This analysis shows the majority of memory accesses are delayed by less than 1 ns, with higher access latencies (caused by resume state reconstruction) subsiding within 0.25 seconds of a system resume. In effect, the MECU provides zero-cost steady state memory confidentiality for non-volatile main memory.

Users commonly download, patch, and use applications such as email clients, office applications, and media-players from the Internet. Such applications are run with the user's full permissions. Because system protections do not differentiate applications, any malcode present in the downloaded software can compromise or otherwise leak all user data. Interestingly, our investigations indicate that common applications often adhere to recognizable workflows on user data. In this paper, we take advantage of this reality by developing protection mechanisms that "pin'' user files to the applications that may use them. These mechanisms restrict access to user data to explicitly stated workflows--thus preventing malcode from exploiting user data not associated with that application. We describe our implementation of PinUP on the Linux Security Modules framework, explore its performance, and study several practical use cases. Through these activities, we show that user data can be protected from untrusted applications while retaining the ability to receive the benefits of those applications.

Rootkits are now prevalent in the wild. Users affected by rootkits are subject to the abuse of their data and resources, often unknowingly. Suchmalware becomes even more dangerous when it is persistent-infected disk images allow the malware to exist across reboots and prevent patches or system repairs from being successfully applied. In this paper, we introduce rootkit-resistant disks (RRD) that label all immutable system binaries and configuration files at installation time. During normal operation, the disk controller inspects all write operations received from the host operating system and denies those made for labeled blocks. To upgrade, the host is booted into a safe state and system blocks can only be modified if a security token is attached to the disk controller. By enforcing immutability at the disk controller, we prevent a compromised operating system from infecting its on-disk image.

Cellular networks are a critical component of the economic and social infrastructures in which we live. In addition to voice services, these networks deliver alphanumeric text messages to the vast majority of wireless subscribers. To encourage the expansion of this new service, telecommunications c ompanies offer connections between their networks and the Internet. The ramifications of such connections, however, have not been fully recognized. In this paper, we evaluate the security impact of the SMS interface on the availability of the cellular phone network. Specifically, we describe the ability to deny voice service to cities the size of Washington DC and Manhattan with little more than a cable modem. Moreover, attacks targeting the entire United States are feasible with resources available to medium-sized zombie networks. This analysis begins with an exploration of the structure of cellular networks. We then characterize network behavior and explore a number of reconnaissance techniques aimed at effectively targeting attacks on these systems. We conclude by discussing countermeasures that mitigate or eliminate the threats introduced by these attacks.

Determining whether a user or system is exercising appropriate security practices is difficult in any context. Such difficulties are particularly pronounced when uncontrolled or unknown platforms join public networks. Commonly practiced techniques used to vet these hosts, such as system scans, have the potential to infringe on the privacy of users. In this article, we show that it is possible for clients to prove both the presence and proper functioning of security infrastructure without allowing unrestricted access to their system. We demonstrate this approach, specifically applied to antivirus security, by requiring clients seeking admission to a network to positively identify the presence or absence of malcode in a series of puzzles. The implementation of this mechanism and its application to real networks are also explored. In so doing, we demonstrate that it is not necessary for an administrator to be invasive to determine whether a client implements required security practices.

THE proliferation of computers in society has meant that organizational and personal assets are increasingly stored and manipulated by software systems. The scale of misuse of these assets has also increased because of their worldwide accessibility through the Internet and the automation of systems. Security is concerned with the prevention of such misuse. While no system can be made completely secure, understanding the context in which a system will be deployed and used, the risks and threats of its misuse, and the systematic development of its software are increasingly recognized as critical to its success. The cross-fertilization of systems development techniques from software engineering and security engineering offers opportunities to minimize duplication of research efforts in both areas and, more importantly, to bridge gaps in our knowledge of how to develop secure softwareintensive systems. The aim of this special issue is to publish novel research work that draws upon software engineering to develop secure systems more effectively. Its scope covers the processes, techniques, technology, people, and knowledge bases that have, or need, the capability to contribute to producing more secure software-intensive systems. In response to the call for papers for this special section, we received 41 submissions, regarding software engineering issues addressing the requirements, design, coding, testing, and maintenance of secure software systems. Each paper was reviewed by at least three expert referees. After two rounds of reviewing, we selected six papers which focus on requirements and design of secure software. The first two papers address both security and privacy requirements, making use of varying degrees of formalism to represent and analyze those requirements. “Analyzing Regulatory Rules for Privacy and Security Requirements” by Travis Breaux and Annie Anton addresses the often overwhelming complexity of regulatory requirements of financial, healthcare, and other software. The formalism and process presented glean enforceable security policy directly from the regulatory statutes. “Privately Finding Specifications” by Westley Weimer and Nina Mishra deals with one of the daunting realities of data sharing, the fact that it is often an all or nothing proposition. This paper describes an attempt to mitigate oversharing in the discovery of software specifications by perturbing program traces. The careful addition of noise into the traces allows specification discovery while preventing the exposure of other sensitive aspects of the program. The next three papers consider how the design of software systems can be realized through security infrastructure. “Semantics-Based Design for Secure Web Services” by Massimo Bartoletti, Pierpaolo Degano, Gian Luigi Ferrari, and Roberto Zunino considers how to develop secure Web services by formally reasoning about policy compliance over historical behaviors. In essence, Web services “contract” (compose) with those systems that respect policies of interest, thereby ensuring globally secure behavior. “Provable Protection against Web Application Vulnerabilities Related to Session Data Dependencies” by Lieven Desmet, Pierre Verbaeten, Wouter Joosen, and Frank Piessens acknowledge recent advances in secure Web application design and development that have made online systems safer. The techniques detailed in this paper prevent misuse of often loosely coupled session dependencies in and among Web applications. “WASP: Protecting Web Applications Using Positive Tainting and Syntax-Aware Evaluation” by William Halfond, Alessandro Orso, and Panagiotis Manolios presents a novel method for preventing SQL injection attacks—attacks in which the adversary inserts arbitrary database query code into an application by manipulating input strings. The paper uses language techniques to dynamically annotate ”trusted” strings, thereby avoiding any use of potentially unsafe strings. The final paper presents a method of certifying that a software system meets its security requirements. “Applying Formal Methods to a Certifiably Secure Software System” by Connie Heitmeyer, Myla Archer, Elizabeth Leonard, and John McLean adds to recent advances that are beginning to make this costly and complex process of formal verification tractable. This paper presents a novel certification method that uses formalized security models to construct a mechanized proof of security over a real-world target system. The papers in this special section demonstrate the strength of research in the area of engineering secure software. If the range and strength of the submissions to the special section are anything to go by, the area is healthy and vibrant and we fully expect many of the submissions that IEEE TRANSACTIONS ON SOFTWARE ENGINEERING, VOL. 34, NO. 1, JANUARY/FEBRUARY 2008 3

Integrity measurements provide a means by which distributed systems can assess the trustability of potentially compromised remote hosts. However, current measurement techniques simply assert the identity of software, but provide no indication of the ongoing status of the system or its data. As a result, a number of significant vulnerabilities can result if the system is not configured and managed carefully. To improve the management of a system's integrity, we propose a Root of Trust Installation (ROTI) as a foundation for high integrity systems. A ROTI is a trusted system installer that also asserts the integrity of the trusted computing base software and data that it installs to enable straightforward, comprehensive integrity verification for a system. The ROTI addresses a historically limiting problem in integrity measurement: determining what constitutes a trusted system state in a heterogeneous, evolving environment. Using the ROTI, a high integrity system state is defined by its installer, thus enabling a remote party to verify integrity guarantees that approximate classical integrity models (e.g., Biba). In this paper, we examine what is necessary to prove the integrity of the trusted computing base (sCore) of a distributed security architecture, called the Shamon. We describe the design and implementation of our custom ROTI sCore installer and study the costs and effectiveness of binding system integrity to installation in the distributed Shamon. This demonstration shows that strong integrity guarantees can be efficiently achieved in large, diverse environments with limited administrative overhead.

Security-typed languages (STLs) are powerful tools for provably implementing policy in applications. The pro- grammer maps policy onto programs by annotating types with information flow labels, and the STL compiler guaran- tees that data always obeys its label as it flows within an application. As data flows into or out of an application, however, a runtime system is needed to mediate between the information flow world within the application and the non-information flow world of the operating system. In the few existing STL applications, this problem has been han- dled in ad hoc ways that hindered software engineering and security analysis. In this paper, we present a principled ap- proach to STL runtime system development along with pol- icy infrastructure and class abstractions for the STL, Jif, that implement these principles. We demonstrate the ef- fectiveness of our approach by using our infrastructure to develop a firewall application, FLOWWALL, that provably enforces its policy.

ASes in inter-domain routing receive little information about the quality of the routes they receive. This lack of information can lead to inefficient and even incorrect routing. In this paper, we quantitatively characterize BGP announcements that violate the so-called valley- free property-an indicator that universal best practices are not being preserved in the propagation of routes. Our analysis indicates that valley announcements are more pervasive than expected. Approximately ten thousand valley announcements appear every day and involve a substantial number of prefixes. 11 % of provider ASes propagate valley announcements, with a majority of violations happening at intermediate providers. We find that large surges of violating announcements can be attributed to transient configuration errors. We further propose a dynamic mechanism that provides route propagation information as transitive attributes of BGP. This information implicitly reflects the policies of the ASes along the path, without revealing the relationship of each AS pair. BGP-speaking routers use this information to identify (and presumably avoid) routes that violate the valley-free property.

One practical limitation of structured peer-to-peer (P2P) networks is that they are frequently subject to Sybil attacks: malicious parties can compromise the network by generating and controlling large numbers of shadow identities. In this paper, we propose an admission control system that mitigates Sybil attacks by adaptively constructing a hierarchy of cooperative peers. The admission control system vets joining nodes via client puzzles. A node wishing to join the network is serially challenged by the nodes from a leaf to the root of the hierarchy. Nodes completing the puzzles of all nodes in the chain are provided a cryptographic proof of the vetted identity. We evaluate our solution and show that an adversary must perform days or weeks of effort to obtain even a small percentage of nodes in small P2P networks, and that this effort increases linearly with the size of the network. We further show that we can place a ceiling on the number of IDs any adversary may obtain by requiring periodic reassertion of the IDs continued validity.

The Internet has grown tremendously in terms of the number of users who rely on it and the number of organizations that are connected to it. Characterizing how this growth affects its structure and topology is vitally important to determine the fundamental characteristics and limitations that must be handled, such as address space exhaustion; understanding the process of allocating and delegating address space can help to answer these questions. In this paper, we analyze BGP routing data to study the structure and growth of IPv4 address space allocation, fragmentation and usage. We explore the notion of delegation relationships among prefixes and use this information to construct an autonomous system (AS) delegation tree. We show that delegation in the Internet is not significantly correlated to the underlying topology or AS customer-provider relationships. We also analyze the fragmentation and usage of address space over a period of five years and examine prefixes that are delegated by organizations vs. those that are not delegated. We notice that the address space usage due to delegating prefixes is increasing at the same rate as the address space usage due to non-delegating prefixes. This indicates that fragmentation rate of the address space is actually almost a constant with respect to total address usage. Additionally, we show that most delegation is performed by a small number of organizations, which may aid in the implementation of a public-key infrastructure for the Internet.

The SELinux mandatory access control (MAC) policy has recently added a multi-level security (MLS) model which is able to express a fine granularity of control over a subject's access rights. The problem is that the richness of this policy makes it impractical to verify, by hand, that a given policy has certain important information flow properties or is compliant with another policy. To address this, we have modeled the SELinux MLS policy using a logical specification and implemented that specification in the Prolog language. Furthermore, we have developed some analyses for testing the properties of a given policy as well an algorithm to determine whether one policy is compliant with another. We have implemented these analyses in Prolog and compiled our implementation into a tool for SELinux MLS policy analysis, called PALMS. Using PALMS, we verified some important properties of the SELinux MLS reference policy, namely that it satisfies the simple security condition and *-property defined by Bell and LaPadula [2].

Security-typed languages are an evolving tool for implementing systems with provable security guarantees. However, to date, these tools have only been used to build simple "toy" programs. As described in this paper, we have developed the first real-world, security-typed application: a secure email system written in the Java language variant Jif. Real-world policies are mapped onto the information flows controlled by the language primitives, and we consider the process and tractability of broadly enforcing security policy in commodity applications. We find that while the language provided the rudimentary tools to achieve low-level security goals, additional tools, services, and language extensions were necessary to formulate and enforce application policy. We detail the design and use of these tools. We also show how the strong guarantees of Jif in conjunction with our policy tools can be used to evaluate security. This work serves as a starting point-we have demonstrated that it is possible to implement real-world systems and policy using security-typed languages. However, further investigation of the developer tools and supporting policy infrastructure is necessary before they can fulfil their considerable promise of enabling more secure systems

The Border Gateway Protocol (BGP) is the de facto interdomain routing protocol on the Internet. While the serious vulnerabilities of BGP are well known, no security solution has been widely deployed. The lack of adoption is largely caused by a failure to find a balance between deployability, cost, and security. In this paper, we consider the design and performance of BGP path authentication constructions that limit resource costs by exploiting route stability. Based on a year-long study of BGP traffic and indirectly supported by findings within the networking community, we observe that routing paths are highly stable. This observation leads to comprehensive and efficient constructions for path authentication. We empirically analyze the resource consumption of the proposed constructions via trace-based simulations. This latter study indicates that our constructions can reduce validation costs by as much as 97.3% over existing proposals while requiring nominal storage resources. We conclude by considering operational issues related to incremental deployment of our solution.

Attributes define, classify, or annotate the datum to which they are assigned. However, traditional attribute architectures and cryptosystems are ill-equipped to provide security in the face of diverse access requirements and environments. In this paper, we introduce a novel secure information management architecture based on emerging attribute-based encryption (ABE) primitives. A policy system that meets the needs of complex policies is defined and illustrated. Based on the needs of those policies, we propose cryptographic optimizations that vastly improve enforcement efficiency. We further explore the use of such policies in two example applications: a HIPAA compliant distributed file system and a social network. A performance analysis of our ABE system and example applications demonstrates the ability to reduce cryptographic costs by as much as 98% over previously proposed constructions. Through this, we demonstrate that our attribute system is an efficient solution for securely managing information in large, loosely-coupled, distributed systems.

The stability and robustness of BGP remains one of the most critical elements in sustaining today’s Internet. In this paper, we study the structure and stability of origin advertisements in inter-domain routing. We visualize and quantitatively characterize the frequency, size, and effect of address assignment and origin changes by analyzing realworld BGP updates for a period of one year from multiple vantage points. Broad classes of prefix behaviors are developed. We show that a significant portion of BGP traffic is due to prefix flapping and explore the contributing factors which include a number of prefixes with abnormal short upand- down cycles. A significant portion of prefixes have high origin stability. Most ASes are involved in few, if any, prefix movement events, while a small number of ASes are responsible for most of the origin churn. Additionally, we find that a high volume of new prefixes can be attributed to actively evolving countries, that some abnormal prefix flapping is most likely due to misconfiguration, and that some culprit ASes characterize the places where multi-origin prefixes oscillate.

The transformation of telecommunications networks from homogeneous closed systems providing only voice services to Internet-connected open networks that provide voice and data services presents significant security challenges. For example, recent research illustrated that a carefully crafted DoS attack via text messaging could incapacitate all voice communications in a metropolitan area with little more than a cable modem. This attack highlights a growing threat to these systems; namely, cellular networks are increasingly exposed to adversaries both in and outside the network. In this paper, we use a combination of modeling and simulation to demonstrate the feasibility of targeted text messaging attacks. Under realistic network conditions, we show that adversaries can achieve blocking rates of more than 70% with only limited resources. We then develop and characterize five techniques from within two broad classes of countermeasures - queue management and resource provisioning. Our analysis demonstrates that these techniques can eliminate or extensively mitigate even the most intense targeted text messaging attacks. We conclude by considering the tradeoffs inherent to the application of these techniques in current and next generation telecommunications networks.

Prior works in communication security policy have focused on general-purpose policy languages and evaluation algorithms. However, because the supporting frameworks often defer enforcement, the correctness of a realization of these policies in softwar

A security policy specifies session participant requirements. However, existing frameworks provide limited facilities for the automated reconciliation of participant policies. This paper considers the limits and methods of reconciliation in a general-purpose policy model. We identify an algorithm for efficient two-policy reconciliation and show that, in the worst-case, reconciliation of three or more policies is intractable. Further, we suggest efficient heuristics for the detection and resolution of intractable reconciliation. Based upon the policy model, we describe the design and implementation of the Ismene policy language. The expressiveness of Ismene, and indirectly of our model, is demonstrated through the representation and exposition of policies supported by existing policy languages. We conclude with brief notes on the integration and enforcement of Ismene policy within the Antigone communication system.

IP networks fundamentally rely on the address resolution protocol (ARP) for proper operation. Unfortunately, vulnerabilities in the ARP protocol enable a raft of IP-based impersonation, man-in-the-middle, or DoS attacks. Proposed countermeasures to these vulnerabilities have yet to simultaneously address backward compatibility and cost requirements. This paper introduces the ticket-based address resolution protocol (TARP). TARP implements security by distributing centrally issued secure MAC/IP address mapping attestations through existing ARP messages. We detail the TARP protocol and its implementation within the Linux operating system. Our experimental analysis shows that TARP improves the costs of implementing ARP security by as much as two orders of magnitude over existing protocols. We conclude by exploring a range of operational issues associated with deploying and administering ARP security.

Cellular networks are a critical component of the economic and social infrastructures in which we live. In addition to voice services, these networks deliver alphanumeric text messages to the vast majority of wireless subscribers. To encourage the expansion of this new service, telecommunications companies offer connections between their networks and the Internet. The ramifications of such connections, however, have not been fully recognized. In this paper, we evaluate the security impact of the SMS interface on the availability of the cellular phone network. Specifically, we demonstrate the ability to deny voice service to cities the size of Washington D.C. and Manhattan with little more than a cable modem. Moreover, attacks targeting the entire United States are feasible with resources available to medium-sized zombie networks. This analysis begins with an exploration of the structure of cellular networks. We then characterize network behavior and explore a number of reconnaissance techniques aimed at effectively targeting attacks on these systems. We conclude by discussing countermeasures that mitigate or eliminate the threats introduced by these attacks.

Nodes forward data on behalf of each other in mobile ad hoc networks. In a civilian application, nodes are assumed to be selfish and rational, i.e., they pursue their own self-interest. Hence, the ability to accurately measure traffic forwarding is critical to ensure proper network operation. These measurements are often used to credit nodes based on their level of participation, or to detect loss. Past solutions employ neighbor monitoring and reporting on node forwarding traffic. These methods are not applicable in civilian networks where neighbor nodes lack the desire or ability to perform the monitoring function. Such environments occur frequently in which neighbor hosts are resource constrained, or in networks where directional antennas are used and reliable monitoring is difficult or impossible. In this paper, we propose a protocol that uses nodes on the data path to securely produce packet forwarding reports. Reporting nodes are chosen randomly and secretly so that malicious nodes cannot modify their behavior based upon the monitoring point. The integrity and authenticity of reports are preserved through the use of secure link layer acknowledgments and monitoring reports. The robustness of the reporting mechanism is strengthened by forwarding the report to multiple destinations (source and destination). We explore the security, cost, and accuracy of our protocol.

Attacks against Internet routing are increasing in number and severity. Contributing greatly to these attacks is the absence of origin authentication: there is no way to validate claims of address ownership or location. The lack of such services enables not only attacks by malicious entities, but indirectly allow seemingly inconsequential miconfigurations to disrupt large portions of the Internet. This paper considers the semantics, design, and costs of origin authentication in interdomain routing. We formalize the semantics of address delegation and use on the Internet, and develop and characterize broad classes of origin authentication proof systems. We estimate the address delegation graph representing the current use of IPv4 address space using available routing data. This effort reveals that current address delegation is dense and relatively static: as few as 16 entities perform 80% of the delegation on the Internet. We conclude by evaluating the proposed services via traced based simulation. Our simulation shows the enhanced proof systems can significantly reduce resource costs associated with origin authentication.

Forward-secure signatures (FSSs) have recently received much attention from the cryptographic theory community as a potentially realistic way to mitigate many of the difficulties digital signatures face with key exposure. However, no previous works have explored the practical performance of these proposed constructions in real-world applications, nor have they compared FSS to traditional, non-forward-secure, signatures in a non-asymptotic way.We present an empirical evaluation of several FSS schemes that looks at the relative performance among different types of FSS as well as between FSS and traditional signatures. Our study provides the following contributions: first, a new methodology for comparing the performance of signature schemes, and second, a thorough examination of the practical performance of FSS. We show that for many cases the best FSS scheme has essentially identical performance to traditional schemes, and even in the worst case is only 2-4 times slower. On the other hand, we also show that if the wrong FSS configuration is used, the performance can be orders of magnitude slower. Our methodology provides a way to prevent such misconfigurations, and we examine common applications of digital signatures using it.We conclude that not only are forward-secure signatures a useful theoretical construct as previous works have shown, but they are also, when used correctly, a very practical solution to some of the problems associated with key exposure in real-world applications. Through our metrics and our reference implementation we provide the tools necessary for developers to efficiently use FSS.

Unauthorized copying of movies is a major concern for the motion picture industry. While unauthorized copies of movies have been distributed via portable physical media for some time, low-cost, high-bandwidth Internet connections and peer-to-peer file sharing networks provide highly efficient distribution media. Many movies are showing up on file sharing networks shortly after, and in some cases prior to, theatrical release. It has been argued that the availability of unauthorized copies directly affects theater attendance and DVD sales, and hence represents a major financial threat to the movie industry. Our research attempts to determine the source of unauthorized copies by studying the availability and characteristics of recent popular movies in file sharing networks. We developed a data set of 312 popular movies and located one or more samples of 183 of these movies on file sharing networks, for a total of 285 movie samples. 77% of these samples appear to have been leaked by industry insiders. Most of our samples appeared on file sharing networks prior to their official consumer DVD release date. Indeed, of the movies that had been released on DVD as of the time of our study, only 5% first appeared after their DVD release date on a web site that indexes file sharing networks, indicating that consumer DVD copying currently represents a relatively minor factor compared with insider leaks. We perform a brief analysis of the movie production and distribution process and identify potential security vulnerabilities that may lead to unauthorized copies becoming available to those who may wish to redistribute them. Finally, we offer recommendations for reducing security vulnerabilities in the movie production and distribution process.

Authorization policy infrastructures are evolving with the complex environments that they support. However, the requirements and technologies supporting context are not yet well understood. Often implemented as condition functions or predefined attributes, context is used to more precisely control when and how policy is enforced. This paper considers context requirements and services in authorization policy. The properties and security requirements of context evaluation are classified. A key observation gleaned from this classification is the degree to which context functions share common properties. The Antigone Condition Framework (ACF) exploits these commonalities to provide a general purpose condition service and associated API. The prototype ACF design is presented and illustrated, and directions for future work considered.

A security policy is a means by which participant session requirements are specified. However, existing frameworks provide limited facilities for the automated reconciliation of participant policies. This paper considers the limits and methods of reconciliation in a general-purpose policy model. We identify an algorithm for efficient two-policy reconciliation, and show that, in the worst-case, reconciliation of three or more policies is intractable. Further, we suggest efficient heuristics for the detection and resolution of intractable reconciliation. Based upon the policy model, we describe the design and implementation of the Ismene policy language. The expressiveness of Ismene, and indirectly of our model, is demonstrated through the representation and exposition of policies supported by existing policy languages. We conclude with brief notes on the integration and enforcement of Ismene policy within the Antigone communication system.

The advent of electronic commerce and personal communications on the Internet has heightened concern over lack of privacy and security. Network services providing a wide range of security related guarantees are increasingly based on public key certificates. A fundamental problem inhibiting the wide acceptance of existing certificate distribution services is the lack of a scalable certificate revocation mechanism. We argue in this paper that the resource requirements of extant revocation mechanisms place a significant burden on certificate servers and network resources. We propose a novel mechanism called windowed revocation that satisfies the security policies and requirements of existing mechanisms and, at the same time, reduces the burden on certificate servers and network resources. We include a proof of correctness of windowed revocation and analyze worst case performance scenarios.